June Will Be 1 Second Longer

by
in internet on (#9XMS)
story imageIt's a dreaded day for many Internet companies: On June 30, an extra second will be added to the clock, creating the potential to wreak havoc on computer systems not equipped to handle the change. The International Earth Rotation and Reference Systems (IERS) announced an extra second will be added at the end of June to account for a discrepancy between Earth's rotation and the atomic clock. The extra second will be added as the clock strikes midnight universal time, meaning the extra second will come for people in the United States at 8 p.m. EDT. There have been 25 instances since 1972 of an extra second being added.

It's possible that programs not equipped to handle the extra second could have an issue. When the last leap second was added on June 30, 2012, it caused issues with a number of websites, including LinkedIn and Yelp. Mozilla, Reddit and Foursquare all experienced system crashes. Qantas was hard hit, too, with a failure of its check-in system creating flight delays across the network.
Reply 9 comments

A bit embarrassing... software developer for years.... (Score: 1)

by tanuki64@pipedot.org on 2015-05-28 09:06 (#9XP1)

...but I simply don't understand the problem.
It's possible that programs not equipped to handle the extra second could have an issue.
What programs? If the clock is not correct I might have a minor problem with builds. I might recompile more than necessary. So What? But crashes? Especially so boring systems like websites? Possibly "lightly" corrupt databases, yes. Perhaps the order of a few posts mixed up, yes. This should be all. I'd probably have think hard how to crash a program on purpose just because the time is one second off.

Re: A bit embarrassing... software developer for years.... (Score: 1, Insightful)

by Anonymous Coward on 2015-05-28 09:37 (#9XQV)

What programs?
Shit programs!

Time should always be stored and processed as a timestamp. Human readable date strings are meaningless to computers and should be used for display only.

Re: A bit embarrassing... software developer for years.... (Score: 0)

by billshooterofbul@pipedot.org on 2015-05-29 20:10 (#A1C0)

Sadly, there are more out there than their should be. The consequences of such issues should be relatively minor. Maybe the program will crash or display something wrong. It probably doesn't matter fro most apps.

Re: A bit embarrassing... software developer for years.... (Score: 3, Funny)

by evilviper@pipedot.org on 2015-05-29 21:19 (#A1G5)

Sadly, there are more duplicate posts from billshooterofbul out there than their should be.

If ANYONE ever deserved the "Redundant" mod...

Re: A bit embarrassing... software developer for years.... (Score: 1)

by bryan@pipedot.org on 2015-06-02 21:32 (#A9XW)

Sorry, my fault. Language translation temporarilly dissabled again.

Re: A bit embarrassing... software developer for years.... (Score: 2, Interesting)

by seriously@pipedot.org on 2015-05-28 10:12 (#9XSV)

I remember reading that technical post on Google blog about problems they had with leap seconds and how they fixed them.

http://googleblog.blogspot.be/2011/09/time-technology-and-leaping-seconds.html

The section "Why time matters at Google" might especially be of interest ;)

Re: A bit embarrassing... software developer for years.... (Score: 1)

by tanuki64@pipedot.org on 2015-05-28 10:21 (#9XTX)

Exactly what I expected: Minor problems with consistency: "Does email that comes in during that second get stored correctly?". Ok, for a service like Google this might not be a minor problem. But this is light years away from a system crash.

Re: A bit embarrassing... software developer for years.... (Score: 1)

by seriously@pipedot.org on 2015-05-28 11:57 (#9Y0H)

Mmmh, I agree, in Google's case it seems more about synchronisation of their multiple machines across their multiple datacenters (and locking issues that go with it). But a bit of googling showed some other more crash-like reports:

http://blog.fastmail.com/2012/07/03/a-story-of-leaping-seconds/
http://lkml.iu.edu/hypermail/linux/kernel/1203.1/04598.html
https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=6b43ae8a619d17c4935c3320d2ef9e92bdeed05d
https://access.redhat.com/solutions/154713

In this case, the "2012 crash", it was kernel-related (something about a multi-CPU race inside the kernel that caused lock issues if I get this right), so not really a problem that you could do anything to avoid in userland.

Now if you're using *BSD, of course, you're most certainly safe ;)

Re: A bit embarrassing... software developer for years.... (Score: 0)

by Anonymous Coward on 2015-05-28 15:52 (#9YEF)

Thanks, I hadn't read that and it appears that the problem is:
Computers traditionally accommodate leap seconds by setting their clock backwards by one second at the very end of the day
The "Time moved backwards by more than 1 second" error from dovecot now makes perfect sense. That this can lead to locking and concurrency issues demonstrates leap seconds are incorrectly implemented. As we are no longer in Nineteen-Seventy-Fucking-Nine; I stand by my original point!